iSixSigma

monte carlo simulation to get 95% confidence and prediction intervals

Six Sigma – iSixSigma Forums General Forums Tools & Templates monte carlo simulation to get 95% confidence and prediction intervals

This topic contains 11 replies, has 6 voices, and was last updated by  MBBinWI 9 years, 2 months ago.

Viewing 12 posts - 1 through 12 (of 12 total)
  • Author
    Posts
  • #53535

    c
    Member

    How do I use monte carlo simulation to get 95% confidence and prediction intervals for my model in linear and multiple regression analysis ?

    0
    #190509

    Robert Butler
    Participant

    If you build a regression model you already have an estimate of the confidence intervals for both the overall model as well as the coefficents.

    0
    #190511

    c
    Member

    yes i know. i mean how do i make use of simulation techniques using monte carlo to get the 95% CI and PI for observation given X1, X2, X3 = given values.

    currently i am using statsgraphics but i would like to try out using monte carlo. what are the steps in Minitab using monte carlo ?

    0
    #190514

    Jones
    Participant

    Though I’ve never used it, Minitab produced a article on the topic. You can find it on Minitab’s website here:
    [url=http://www.minitab.com/uploadedFiles/Shared_Resources/Documents/Articles/Monte%20Carlo%20Simulation%]

    Hope this helps.

    0
    #190522

    Gadgil
    Member

    You can build the model using Crystal Ball.You will have data sets for X’s from which you have arrived at the regression equation.For each X data set,identify the data distribution(Green cells-Definition cells) in Crystal Ball.For Y cell put the regression equation in the cell & make it Forecast cell(Blue cell in Crystal Ball).When run the simulation.It takes all the values from the data sets of Xs depending upon the distribution & then plots a prediction grpah for Y.You can play with graph by moving the pointer & having values for different certainty %.

    0
    #190528

    c
    Member

    hi JHJ, the URL returned a message : Bad Request (Invalid URL).

    then i tried
    http://www.minitab.com/uploadedFiles/Shared_Resources/Documents/Articles
    and it states page not found.

    0
    #190529

    c
    Member

    Thanks Vinayak. But i do not have crystal ball.
    anyone familiar with monte carlo using minitab ?

    0
    #190530

    Jones
    Participant

    I’ll try again.

    I think the posting interface limits the amount of characters. You can try to copy and paste the following link into your browser.

    http://www.minitab.com/uploadedFiles/Shared_Resources/Documents/Articles/Monte Carlo Simulation Article.pdf

    If you see “%20” in the post, be sure to replace all instances with a single space.

    0
    #190531

    Jones
    Participant
    #190545

    Paulonis
    Participant

    It does not appear that the Minitab article takes into account the uncertainty of the model in predicting the uncertainty of the model output. The article is about performing a sensitivity analysis assuming a perfect model and uncertain inputs. I don’t think it is what the original poster is looking for. The article does, however, describe an approach for using Minitab for a Monte Carlo analysis.

    To the original poster, what is the purpose of your question? If you’ve run regression in Minitab, you can analytically generate confidence and prediction intervals for any set of X’s. Why would you want to use a Monte Carlo method?

    0
    #190569

    c
    Member

    i am actually comparing the simulation results taken from statsgraphics to results from using Monte Carlo in Minitab.

    I was told that simulation results are done separately for beta0, beta1 etc in the regression equation. However, in monte carlo in mintab, it is able to handle all betas siimultaneously. and thus i would like to try if this is true, and how much the results differ and will there be any visible impact if i have done my analysis in diff softwares.

    0
    #190589

    MBBinWI
    Participant

    Then call up Minitab support and talk to their techies. They are very responsive to enquiries on how analyses are being performed.

    0
Viewing 12 posts - 1 through 12 (of 12 total)

You must be logged in to reply to this topic.